Comparative Analysis of Churn Predictive Models and Factor Identification in Telecom Industry

Abstract

Continual advancement in technology has led an initiative to the competitive environment among the institutes relating to the technological domain. The telecommunication industry is no exception in such cases. There exists immense competition among the telecom service providers for maximization of profit and expansion of market interest by attracting new clients. However, the retention of existing customers is easier and cheaper than acquiring new ones. As the customers are more concerned about the quality of services provided by the institutions it becomes challenging for companies to maintain client satisfaction. The CRM as well as analysts need to recognize the potential churners and the cause of their migration. This paper suggests a framework that employs machine learning and deep learning techniques for determining churn customers as well as distinguishes notable factors that typically govern the customer towards churn. Firstly, the classification between churn and non-churn customers is conducted utilizing both machine learning and deep learning algorithms where Random Forest achieved supremacy over others and followed by the deep learning models CNN and MLP. Besides the work deduced the significant factors affecting the churning procedure by applying Attribute Selection Techniques. The experimentation results unveil the prediction models that recognize the potential churners with optimal accuracy and the important factors that show impact over the churning of the customer. The findings acquired from this research are hoped to be lucrative for the companies in the present world for taking an effective decision and acting accurately in terms of customer retention.